ih0dl / octopus

A coding assistant with a 32k context window.

27 Pulls Updated 4 months ago

4 Tags

octopus ... /

model

90038122ded1 · 24GB

Metadata

general.architecture

llama
general.file_type

7
general.name

Codestral-22B-v0.1
general.quantization_version

2
llama.attention.head_count

48
llama.attention.head_count_kv

8
llama.attention.layer_norm_rms_epsilon

1e-05
llama.block_count

56
llama.context_length

32768
llama.embedding_length

6144
llama.feed_forward_length

16384
llama.rope.dimension_count

128
llama.rope.freq_base

1e+06
llama.vocab_size

32768
tokenizer.ggml.add_bos_token

true
tokenizer.ggml.add_eos_token

false
tokenizer.ggml.add_space_prefix

true
tokenizer.ggml.bos_token_id

1
tokenizer.ggml.eos_token_id

2
tokenizer.ggml.model

llama
tokenizer.ggml.pre

default
tokenizer.ggml.scores

[0, 0, 0, 0, 0, ...]
tokenizer.ggml.token_type

[2, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens

[<unk>, <s>, </s>, [INST], [/INST], ...]
tokenizer.ggml.unknown_token_id

0

Tensor

Name

Type

Shape
token_embd.weight

Q8_0

[6144, 32768]

blk.0

blk.0.attn_k.weight

Q8_0

[6144, 1024]
blk.0.attn_norm.weight

F32

[6144]
blk.0.attn_output.weight

Q8_0

[6144, 6144]
blk.0.attn_q.weight

Q8_0

[6144, 6144]
blk.0.attn_v.weight

Q8_0

[6144, 1024]
blk.0.ffn_down.weight

Q8_0

[16384, 6144]
blk.0.ffn_gate.weight

Q8_0

[6144, 16384]
blk.0.ffn_norm.weight

F32

[6144]
blk.0.ffn_up.weight

Q8_0

[6144, 16384]

blk.1

blk.1.attn_k.weight

Q8_0

[6144, 1024]
blk.1.attn_norm.weight

F32

[6144]
blk.1.attn_output.weight

Q8_0

[6144, 6144]
blk.1.attn_q.weight

Q8_0

[6144, 6144]
blk.1.attn_v.weight

Q8_0

[6144, 1024]
blk.1.ffn_down.weight

Q8_0

[16384, 6144]
blk.1.ffn_gate.weight

Q8_0

[6144, 16384]
blk.1.ffn_norm.weight

F32

[6144]
blk.1.ffn_up.weight

Q8_0

[6144, 16384]

blk.2

blk.2.attn_k.weight

Q8_0

[6144, 1024]
blk.2.attn_norm.weight

F32

[6144]
blk.2.attn_output.weight

Q8_0

[6144, 6144]
blk.2.attn_q.weight

Q8_0

[6144, 6144]
blk.2.attn_v.weight

Q8_0

[6144, 1024]
blk.2.ffn_down.weight

Q8_0

[16384, 6144]
blk.2.ffn_gate.weight

Q8_0

[6144, 16384]
blk.2.ffn_norm.weight

F32

[6144]
blk.2.ffn_up.weight

Q8_0

[6144, 16384]

blk.3

blk.3.attn_k.weight

Q8_0

[6144, 1024]
blk.3.attn_norm.weight

F32

[6144]
blk.3.attn_output.weight

Q8_0

[6144, 6144]
blk.3.attn_q.weight

Q8_0

[6144, 6144]
blk.3.attn_v.weight

Q8_0

[6144, 1024]
blk.3.ffn_down.weight

Q8_0

[16384, 6144]
blk.3.ffn_gate.weight

Q8_0

[6144, 16384]
blk.3.ffn_norm.weight

F32

[6144]
blk.3.ffn_up.weight

Q8_0

[6144, 16384]

blk.4

blk.4.attn_k.weight

Q8_0

[6144, 1024]
blk.4.attn_norm.weight

F32

[6144]
blk.4.attn_output.weight

Q8_0

[6144, 6144]
blk.4.attn_q.weight

Q8_0

[6144, 6144]
blk.4.attn_v.weight

Q8_0

[6144, 1024]
blk.4.ffn_down.weight

Q8_0

[16384, 6144]
blk.4.ffn_gate.weight

Q8_0

[6144, 16384]
blk.4.ffn_norm.weight

F32

[6144]
blk.4.ffn_up.weight

Q8_0

[6144, 16384]

blk.5

blk.5.attn_k.weight

Q8_0

[6144, 1024]
blk.5.attn_norm.weight

F32

[6144]
blk.5.attn_output.weight

Q8_0

[6144, 6144]
blk.5.attn_q.weight

Q8_0

[6144, 6144]
blk.5.attn_v.weight

Q8_0

[6144, 1024]
blk.5.ffn_down.weight

Q8_0

[16384, 6144]
blk.5.ffn_gate.weight

Q8_0

[6144, 16384]
blk.5.ffn_norm.weight

F32

[6144]
blk.5.ffn_up.weight

Q8_0

[6144, 16384]

blk.6

blk.6.attn_k.weight

Q8_0

[6144, 1024]
blk.6.attn_norm.weight

F32

[6144]
blk.6.attn_output.weight

Q8_0

[6144, 6144]
blk.6.attn_q.weight

Q8_0

[6144, 6144]
blk.6.attn_v.weight

Q8_0

[6144, 1024]
blk.6.ffn_down.weight

Q8_0

[16384, 6144]
blk.6.ffn_gate.weight

Q8_0

[6144, 16384]
blk.6.ffn_norm.weight

F32

[6144]
blk.6.ffn_up.weight

Q8_0

[6144, 16384]

blk.7

blk.7.attn_k.weight

Q8_0

[6144, 1024]
blk.7.attn_norm.weight

F32

[6144]
blk.7.attn_output.weight

Q8_0

[6144, 6144]
blk.7.attn_q.weight

Q8_0

[6144, 6144]
blk.7.attn_v.weight

Q8_0

[6144, 1024]
blk.7.ffn_down.weight

Q8_0

[16384, 6144]
blk.7.ffn_gate.weight

Q8_0

[6144, 16384]
blk.7.ffn_norm.weight

F32

[6144]
blk.7.ffn_up.weight

Q8_0

[6144, 16384]

blk.8

blk.8.attn_k.weight

Q8_0

[6144, 1024]
blk.8.attn_norm.weight

F32

[6144]
blk.8.attn_output.weight

Q8_0

[6144, 6144]
blk.8.attn_q.weight

Q8_0

[6144, 6144]
blk.8.attn_v.weight

Q8_0

[6144, 1024]
blk.8.ffn_down.weight

Q8_0

[16384, 6144]
blk.8.ffn_gate.weight

Q8_0

[6144, 16384]
blk.8.ffn_norm.weight

F32

[6144]
blk.8.ffn_up.weight

Q8_0

[6144, 16384]

blk.9

blk.9.attn_k.weight

Q8_0

[6144, 1024]
blk.9.attn_norm.weight

F32

[6144]
blk.9.attn_output.weight

Q8_0

[6144, 6144]
blk.9.attn_q.weight

Q8_0

[6144, 6144]
blk.9.attn_v.weight

Q8_0

[6144, 1024]
blk.9.ffn_down.weight

Q8_0

[16384, 6144]
blk.9.ffn_gate.weight

Q8_0

[6144, 16384]
blk.9.ffn_norm.weight

F32

[6144]
blk.9.ffn_up.weight

Q8_0

[6144, 16384]

blk.10

blk.10.attn_k.weight

Q8_0

[6144, 1024]
blk.10.attn_norm.weight

F32

[6144]
blk.10.attn_output.weight

Q8_0

[6144, 6144]
blk.10.attn_q.weight

Q8_0

[6144, 6144]
blk.10.attn_v.weight

Q8_0

[6144, 1024]
blk.10.ffn_down.weight

Q8_0

[16384, 6144]
blk.10.ffn_gate.weight

Q8_0

[6144, 16384]
blk.10.ffn_norm.weight

F32

[6144]
blk.10.ffn_up.weight

Q8_0

[6144, 16384]

blk.11

blk.11.attn_k.weight

Q8_0

[6144, 1024]
blk.11.attn_norm.weight

F32

[6144]
blk.11.attn_output.weight

Q8_0

[6144, 6144]
blk.11.attn_q.weight

Q8_0

[6144, 6144]
blk.11.attn_v.weight

Q8_0

[6144, 1024]
blk.11.ffn_down.weight

Q8_0

[16384, 6144]
blk.11.ffn_gate.weight

Q8_0

[6144, 16384]
blk.11.ffn_norm.weight

F32

[6144]
blk.11.ffn_up.weight

Q8_0

[6144, 16384]

blk.12

blk.12.attn_k.weight

Q8_0

[6144, 1024]
blk.12.attn_norm.weight

F32

[6144]
blk.12.attn_output.weight

Q8_0

[6144, 6144]
blk.12.attn_q.weight

Q8_0

[6144, 6144]
blk.12.attn_v.weight

Q8_0

[6144, 1024]
blk.12.ffn_down.weight

Q8_0

[16384, 6144]
blk.12.ffn_gate.weight

Q8_0

[6144, 16384]
blk.12.ffn_norm.weight

F32

[6144]
blk.12.ffn_up.weight

Q8_0

[6144, 16384]

blk.13

blk.13.attn_k.weight

Q8_0

[6144, 1024]
blk.13.attn_norm.weight

F32

[6144]
blk.13.attn_output.weight

Q8_0

[6144, 6144]
blk.13.attn_q.weight

Q8_0

[6144, 6144]
blk.13.attn_v.weight

Q8_0

[6144, 1024]
blk.13.ffn_down.weight

Q8_0

[16384, 6144]
blk.13.ffn_gate.weight

Q8_0

[6144, 16384]
blk.13.ffn_norm.weight

F32

[6144]
blk.13.ffn_up.weight

Q8_0

[6144, 16384]

blk.14

blk.14.attn_k.weight

Q8_0

[6144, 1024]
blk.14.attn_norm.weight

F32

[6144]
blk.14.attn_output.weight

Q8_0

[6144, 6144]
blk.14.attn_q.weight

Q8_0

[6144, 6144]
blk.14.attn_v.weight

Q8_0

[6144, 1024]
blk.14.ffn_down.weight

Q8_0

[16384, 6144]
blk.14.ffn_gate.weight

Q8_0

[6144, 16384]
blk.14.ffn_norm.weight

F32

[6144]
blk.14.ffn_up.weight

Q8_0

[6144, 16384]

blk.15

blk.15.attn_k.weight

Q8_0

[6144, 1024]
blk.15.attn_norm.weight

F32

[6144]
blk.15.attn_output.weight

Q8_0

[6144, 6144]
blk.15.attn_q.weight

Q8_0

[6144, 6144]
blk.15.attn_v.weight

Q8_0

[6144, 1024]
blk.15.ffn_down.weight

Q8_0

[16384, 6144]
blk.15.ffn_gate.weight

Q8_0

[6144, 16384]
blk.15.ffn_norm.weight

F32

[6144]
blk.15.ffn_up.weight

Q8_0

[6144, 16384]

blk.16

blk.16.attn_k.weight

Q8_0

[6144, 1024]
blk.16.attn_norm.weight

F32

[6144]
blk.16.attn_output.weight

Q8_0

[6144, 6144]
blk.16.attn_q.weight

Q8_0

[6144, 6144]
blk.16.attn_v.weight

Q8_0

[6144, 1024]
blk.16.ffn_down.weight

Q8_0

[16384, 6144]
blk.16.ffn_gate.weight

Q8_0

[6144, 16384]
blk.16.ffn_norm.weight

F32

[6144]
blk.16.ffn_up.weight

Q8_0

[6144, 16384]

blk.17

blk.17.attn_k.weight

Q8_0

[6144, 1024]
blk.17.attn_norm.weight

F32

[6144]
blk.17.attn_output.weight

Q8_0

[6144, 6144]
blk.17.attn_q.weight

Q8_0

[6144, 6144]
blk.17.attn_v.weight

Q8_0

[6144, 1024]
blk.17.ffn_down.weight

Q8_0

[16384, 6144]
blk.17.ffn_gate.weight

Q8_0

[6144, 16384]
blk.17.ffn_norm.weight

F32

[6144]
blk.17.ffn_up.weight

Q8_0

[6144, 16384]

blk.18

blk.18.attn_k.weight

Q8_0

[6144, 1024]
blk.18.attn_norm.weight

F32

[6144]
blk.18.attn_output.weight

Q8_0

[6144, 6144]
blk.18.attn_q.weight

Q8_0

[6144, 6144]
blk.18.attn_v.weight

Q8_0

[6144, 1024]
blk.18.ffn_down.weight

Q8_0

[16384, 6144]
blk.18.ffn_gate.weight

Q8_0

[6144, 16384]
blk.18.ffn_norm.weight

F32

[6144]
blk.18.ffn_up.weight

Q8_0

[6144, 16384]

blk.19

blk.19.attn_k.weight

Q8_0

[6144, 1024]
blk.19.attn_norm.weight

F32

[6144]
blk.19.attn_output.weight

Q8_0

[6144, 6144]
blk.19.attn_q.weight

Q8_0

[6144, 6144]
blk.19.attn_v.weight

Q8_0

[6144, 1024]
blk.19.ffn_down.weight

Q8_0

[16384, 6144]
blk.19.ffn_gate.weight

Q8_0

[6144, 16384]
blk.19.ffn_norm.weight

F32

[6144]
blk.19.ffn_up.weight

Q8_0

[6144, 16384]

blk.20

blk.20.attn_k.weight

Q8_0

[6144, 1024]
blk.20.attn_norm.weight

F32

[6144]
blk.20.attn_output.weight

Q8_0

[6144, 6144]
blk.20.attn_q.weight

Q8_0

[6144, 6144]
blk.20.attn_v.weight

Q8_0

[6144, 1024]
blk.20.ffn_down.weight

Q8_0

[16384, 6144]
blk.20.ffn_gate.weight

Q8_0

[6144, 16384]
blk.20.ffn_norm.weight

F32

[6144]
blk.20.ffn_up.weight

Q8_0

[6144, 16384]

blk.21

blk.21.attn_k.weight

Q8_0

[6144, 1024]
blk.21.attn_norm.weight

F32

[6144]
blk.21.attn_output.weight

Q8_0

[6144, 6144]
blk.21.attn_q.weight

Q8_0

[6144, 6144]
blk.21.attn_v.weight

Q8_0

[6144, 1024]
blk.21.ffn_down.weight

Q8_0

[16384, 6144]
blk.21.ffn_gate.weight

Q8_0

[6144, 16384]
blk.21.ffn_norm.weight

F32

[6144]
blk.21.ffn_up.weight

Q8_0

[6144, 16384]

blk.22

blk.22.attn_k.weight

Q8_0

[6144, 1024]
blk.22.attn_norm.weight

F32

[6144]
blk.22.attn_output.weight

Q8_0

[6144, 6144]
blk.22.attn_q.weight

Q8_0

[6144, 6144]
blk.22.attn_v.weight

Q8_0

[6144, 1024]
blk.22.ffn_down.weight

Q8_0

[16384, 6144]
blk.22.ffn_gate.weight

Q8_0

[6144, 16384]
blk.22.ffn_norm.weight

F32

[6144]
blk.22.ffn_up.weight

Q8_0

[6144, 16384]

blk.23

blk.23.attn_k.weight

Q8_0

[6144, 1024]
blk.23.attn_norm.weight

F32

[6144]
blk.23.attn_output.weight

Q8_0

[6144, 6144]
blk.23.attn_q.weight

Q8_0

[6144, 6144]
blk.23.attn_v.weight

Q8_0

[6144, 1024]
blk.23.ffn_down.weight

Q8_0

[16384, 6144]
blk.23.ffn_gate.weight

Q8_0

[6144, 16384]
blk.23.ffn_norm.weight

F32

[6144]
blk.23.ffn_up.weight

Q8_0

[6144, 16384]

blk.24

blk.24.attn_k.weight

Q8_0

[6144, 1024]
blk.24.attn_norm.weight

F32

[6144]
blk.24.attn_output.weight

Q8_0

[6144, 6144]
blk.24.attn_q.weight

Q8_0

[6144, 6144]
blk.24.attn_v.weight

Q8_0

[6144, 1024]
blk.24.ffn_down.weight

Q8_0

[16384, 6144]
blk.24.ffn_gate.weight

Q8_0

[6144, 16384]
blk.24.ffn_norm.weight

F32

[6144]
blk.24.ffn_up.weight

Q8_0

[6144, 16384]

blk.25

blk.25.attn_k.weight

Q8_0

[6144, 1024]
blk.25.attn_norm.weight

F32

[6144]
blk.25.attn_output.weight

Q8_0

[6144, 6144]
blk.25.attn_q.weight

Q8_0

[6144, 6144]
blk.25.attn_v.weight

Q8_0

[6144, 1024]
blk.25.ffn_down.weight

Q8_0

[16384, 6144]
blk.25.ffn_gate.weight

Q8_0

[6144, 16384]
blk.25.ffn_norm.weight

F32

[6144]
blk.25.ffn_up.weight

Q8_0

[6144, 16384]

blk.26

blk.26.attn_k.weight

Q8_0

[6144, 1024]
blk.26.attn_norm.weight

F32

[6144]
blk.26.attn_output.weight

Q8_0

[6144, 6144]
blk.26.attn_q.weight

Q8_0

[6144, 6144]
blk.26.attn_v.weight

Q8_0

[6144, 1024]
blk.26.ffn_down.weight

Q8_0

[16384, 6144]
blk.26.ffn_gate.weight

Q8_0

[6144, 16384]
blk.26.ffn_norm.weight

F32

[6144]
blk.26.ffn_up.weight

Q8_0

[6144, 16384]

blk.27

blk.27.attn_k.weight

Q8_0

[6144, 1024]
blk.27.attn_norm.weight

F32

[6144]
blk.27.attn_output.weight

Q8_0

[6144, 6144]
blk.27.attn_q.weight

Q8_0

[6144, 6144]
blk.27.attn_v.weight

Q8_0

[6144, 1024]
blk.27.ffn_down.weight

Q8_0

[16384, 6144]
blk.27.ffn_gate.weight

Q8_0

[6144, 16384]
blk.27.ffn_norm.weight

F32

[6144]
blk.27.ffn_up.weight

Q8_0

[6144, 16384]

blk.28

blk.28.attn_k.weight

Q8_0

[6144, 1024]
blk.28.attn_norm.weight

F32

[6144]
blk.28.attn_output.weight

Q8_0

[6144, 6144]
blk.28.attn_q.weight

Q8_0

[6144, 6144]
blk.28.attn_v.weight

Q8_0

[6144, 1024]
blk.28.ffn_down.weight

Q8_0

[16384, 6144]
blk.28.ffn_gate.weight

Q8_0

[6144, 16384]
blk.28.ffn_norm.weight

F32

[6144]
blk.28.ffn_up.weight

Q8_0

[6144, 16384]

blk.29

blk.29.attn_k.weight

Q8_0

[6144, 1024]
blk.29.attn_norm.weight

F32

[6144]
blk.29.attn_output.weight

Q8_0

[6144, 6144]
blk.29.attn_q.weight

Q8_0

[6144, 6144]
blk.29.attn_v.weight

Q8_0

[6144, 1024]
blk.29.ffn_down.weight

Q8_0

[16384, 6144]
blk.29.ffn_gate.weight

Q8_0

[6144, 16384]
blk.29.ffn_norm.weight

F32

[6144]
blk.29.ffn_up.weight

Q8_0

[6144, 16384]

blk.30

blk.30.attn_k.weight

Q8_0

[6144, 1024]
blk.30.attn_norm.weight

F32

[6144]
blk.30.attn_output.weight

Q8_0

[6144, 6144]
blk.30.attn_q.weight

Q8_0

[6144, 6144]
blk.30.attn_v.weight

Q8_0

[6144, 1024]
blk.30.ffn_down.weight

Q8_0

[16384, 6144]
blk.30.ffn_gate.weight

Q8_0

[6144, 16384]
blk.30.ffn_norm.weight

F32

[6144]
blk.30.ffn_up.weight

Q8_0

[6144, 16384]

blk.31

blk.31.attn_k.weight

Q8_0

[6144, 1024]
blk.31.attn_norm.weight

F32

[6144]
blk.31.attn_output.weight

Q8_0

[6144, 6144]
blk.31.attn_q.weight

Q8_0

[6144, 6144]
blk.31.attn_v.weight

Q8_0

[6144, 1024]
blk.31.ffn_down.weight

Q8_0

[16384, 6144]
blk.31.ffn_gate.weight

Q8_0

[6144, 16384]
blk.31.ffn_norm.weight

F32

[6144]
blk.31.ffn_up.weight

Q8_0

[6144, 16384]

blk.32

blk.32.attn_k.weight

Q8_0

[6144, 1024]
blk.32.attn_norm.weight

F32

[6144]
blk.32.attn_output.weight

Q8_0

[6144, 6144]
blk.32.attn_q.weight

Q8_0

[6144, 6144]
blk.32.attn_v.weight

Q8_0

[6144, 1024]
blk.32.ffn_down.weight

Q8_0

[16384, 6144]
blk.32.ffn_gate.weight

Q8_0

[6144, 16384]
blk.32.ffn_norm.weight

F32

[6144]
blk.32.ffn_up.weight

Q8_0

[6144, 16384]

blk.33

blk.33.attn_k.weight

Q8_0

[6144, 1024]
blk.33.attn_norm.weight

F32

[6144]
blk.33.attn_output.weight

Q8_0

[6144, 6144]
blk.33.attn_q.weight

Q8_0

[6144, 6144]
blk.33.attn_v.weight

Q8_0

[6144, 1024]
blk.33.ffn_down.weight

Q8_0

[16384, 6144]
blk.33.ffn_gate.weight

Q8_0

[6144, 16384]
blk.33.ffn_norm.weight

F32

[6144]
blk.33.ffn_up.weight

Q8_0

[6144, 16384]

blk.34

blk.34.attn_k.weight

Q8_0

[6144, 1024]
blk.34.attn_norm.weight

F32

[6144]
blk.34.attn_output.weight

Q8_0

[6144, 6144]
blk.34.attn_q.weight

Q8_0

[6144, 6144]
blk.34.attn_v.weight

Q8_0

[6144, 1024]
blk.34.ffn_down.weight

Q8_0

[16384, 6144]
blk.34.ffn_gate.weight

Q8_0

[6144, 16384]
blk.34.ffn_norm.weight

F32

[6144]
blk.34.ffn_up.weight

Q8_0

[6144, 16384]

blk.35

blk.35.attn_k.weight

Q8_0

[6144, 1024]
blk.35.attn_norm.weight

F32

[6144]
blk.35.attn_output.weight

Q8_0

[6144, 6144]
blk.35.attn_q.weight

Q8_0

[6144, 6144]
blk.35.attn_v.weight

Q8_0

[6144, 1024]
blk.35.ffn_down.weight

Q8_0

[16384, 6144]
blk.35.ffn_gate.weight

Q8_0

[6144, 16384]
blk.35.ffn_norm.weight

F32

[6144]
blk.35.ffn_up.weight

Q8_0

[6144, 16384]

blk.36

blk.36.attn_k.weight

Q8_0

[6144, 1024]
blk.36.attn_norm.weight

F32

[6144]
blk.36.attn_output.weight

Q8_0

[6144, 6144]
blk.36.attn_q.weight

Q8_0

[6144, 6144]
blk.36.attn_v.weight

Q8_0

[6144, 1024]
blk.36.ffn_down.weight

Q8_0

[16384, 6144]
blk.36.ffn_gate.weight

Q8_0

[6144, 16384]
blk.36.ffn_norm.weight

F32

[6144]
blk.36.ffn_up.weight

Q8_0

[6144, 16384]

blk.37

blk.37.attn_k.weight

Q8_0

[6144, 1024]
blk.37.attn_norm.weight

F32

[6144]
blk.37.attn_output.weight

Q8_0

[6144, 6144]
blk.37.attn_q.weight

Q8_0

[6144, 6144]
blk.37.attn_v.weight

Q8_0

[6144, 1024]
blk.37.ffn_down.weight

Q8_0

[16384, 6144]
blk.37.ffn_gate.weight

Q8_0

[6144, 16384]
blk.37.ffn_norm.weight

F32

[6144]
blk.37.ffn_up.weight

Q8_0

[6144, 16384]

blk.38

blk.38.attn_k.weight

Q8_0

[6144, 1024]
blk.38.attn_norm.weight

F32

[6144]
blk.38.attn_output.weight

Q8_0

[6144, 6144]
blk.38.attn_q.weight

Q8_0

[6144, 6144]
blk.38.attn_v.weight

Q8_0

[6144, 1024]
blk.38.ffn_down.weight

Q8_0

[16384, 6144]
blk.38.ffn_gate.weight

Q8_0

[6144, 16384]
blk.38.ffn_norm.weight

F32

[6144]
blk.38.ffn_up.weight

Q8_0

[6144, 16384]

blk.39

blk.39.attn_k.weight

Q8_0

[6144, 1024]
blk.39.attn_norm.weight

F32

[6144]
blk.39.attn_output.weight

Q8_0

[6144, 6144]
blk.39.attn_q.weight

Q8_0

[6144, 6144]
blk.39.attn_v.weight

Q8_0

[6144, 1024]
blk.39.ffn_down.weight

Q8_0

[16384, 6144]
blk.39.ffn_gate.weight

Q8_0

[6144, 16384]
blk.39.ffn_norm.weight

F32

[6144]
blk.39.ffn_up.weight

Q8_0

[6144, 16384]

blk.40

blk.40.attn_k.weight

Q8_0

[6144, 1024]
blk.40.attn_norm.weight

F32

[6144]
blk.40.attn_output.weight

Q8_0

[6144, 6144]
blk.40.attn_q.weight

Q8_0

[6144, 6144]
blk.40.attn_v.weight

Q8_0

[6144, 1024]
blk.40.ffn_down.weight

Q8_0

[16384, 6144]
blk.40.ffn_gate.weight

Q8_0

[6144, 16384]
blk.40.ffn_norm.weight

F32

[6144]
blk.40.ffn_up.weight

Q8_0

[6144, 16384]

blk.41

blk.41.attn_k.weight

Q8_0

[6144, 1024]
blk.41.attn_norm.weight

F32

[6144]
blk.41.attn_output.weight

Q8_0

[6144, 6144]
blk.41.attn_q.weight

Q8_0

[6144, 6144]
blk.41.attn_v.weight

Q8_0

[6144, 1024]
blk.41.ffn_down.weight

Q8_0

[16384, 6144]
blk.41.ffn_gate.weight

Q8_0

[6144, 16384]
blk.41.ffn_norm.weight

F32

[6144]
blk.41.ffn_up.weight

Q8_0

[6144, 16384]

blk.42

blk.42.attn_k.weight

Q8_0

[6144, 1024]
blk.42.attn_norm.weight

F32

[6144]
blk.42.attn_output.weight

Q8_0

[6144, 6144]
blk.42.attn_q.weight

Q8_0

[6144, 6144]
blk.42.attn_v.weight

Q8_0

[6144, 1024]
blk.42.ffn_down.weight

Q8_0

[16384, 6144]
blk.42.ffn_gate.weight

Q8_0

[6144, 16384]
blk.42.ffn_norm.weight

F32

[6144]
blk.42.ffn_up.weight

Q8_0

[6144, 16384]

blk.43

blk.43.attn_k.weight

Q8_0

[6144, 1024]
blk.43.attn_norm.weight

F32

[6144]
blk.43.attn_output.weight

Q8_0

[6144, 6144]
blk.43.attn_q.weight

Q8_0

[6144, 6144]
blk.43.attn_v.weight

Q8_0

[6144, 1024]
blk.43.ffn_down.weight

Q8_0

[16384, 6144]
blk.43.ffn_gate.weight

Q8_0

[6144, 16384]
blk.43.ffn_norm.weight

F32

[6144]
blk.43.ffn_up.weight

Q8_0

[6144, 16384]

blk.44

blk.44.attn_k.weight

Q8_0

[6144, 1024]
blk.44.attn_norm.weight

F32

[6144]
blk.44.attn_output.weight

Q8_0

[6144, 6144]
blk.44.attn_q.weight

Q8_0

[6144, 6144]
blk.44.attn_v.weight

Q8_0

[6144, 1024]
blk.44.ffn_down.weight

Q8_0

[16384, 6144]
blk.44.ffn_gate.weight

Q8_0

[6144, 16384]
blk.44.ffn_norm.weight

F32

[6144]
blk.44.ffn_up.weight

Q8_0

[6144, 16384]

blk.45

blk.45.attn_k.weight

Q8_0

[6144, 1024]
blk.45.attn_norm.weight

F32

[6144]
blk.45.attn_output.weight

Q8_0

[6144, 6144]
blk.45.attn_q.weight

Q8_0

[6144, 6144]
blk.45.attn_v.weight

Q8_0

[6144, 1024]
blk.45.ffn_down.weight

Q8_0

[16384, 6144]
blk.45.ffn_gate.weight

Q8_0

[6144, 16384]
blk.45.ffn_norm.weight

F32

[6144]
blk.45.ffn_up.weight

Q8_0

[6144, 16384]

blk.46

blk.46.attn_k.weight

Q8_0

[6144, 1024]
blk.46.attn_norm.weight

F32

[6144]
blk.46.attn_output.weight

Q8_0

[6144, 6144]
blk.46.attn_q.weight

Q8_0

[6144, 6144]
blk.46.attn_v.weight

Q8_0

[6144, 1024]
blk.46.ffn_down.weight

Q8_0

[16384, 6144]
blk.46.ffn_gate.weight

Q8_0

[6144, 16384]
blk.46.ffn_norm.weight

F32

[6144]
blk.46.ffn_up.weight

Q8_0

[6144, 16384]

blk.47

blk.47.attn_k.weight

Q8_0

[6144, 1024]
blk.47.attn_norm.weight

F32

[6144]
blk.47.attn_output.weight

Q8_0

[6144, 6144]
blk.47.attn_q.weight

Q8_0

[6144, 6144]
blk.47.attn_v.weight

Q8_0

[6144, 1024]
blk.47.ffn_down.weight

Q8_0

[16384, 6144]
blk.47.ffn_gate.weight

Q8_0

[6144, 16384]
blk.47.ffn_norm.weight

F32

[6144]
blk.47.ffn_up.weight

Q8_0

[6144, 16384]

blk.48

blk.48.attn_k.weight

Q8_0

[6144, 1024]
blk.48.attn_norm.weight

F32

[6144]
blk.48.attn_output.weight

Q8_0

[6144, 6144]
blk.48.attn_q.weight

Q8_0

[6144, 6144]
blk.48.attn_v.weight

Q8_0

[6144, 1024]
blk.48.ffn_down.weight

Q8_0

[16384, 6144]
blk.48.ffn_gate.weight

Q8_0

[6144, 16384]
blk.48.ffn_norm.weight

F32

[6144]
blk.48.ffn_up.weight

Q8_0

[6144, 16384]

blk.49

blk.49.attn_k.weight

Q8_0

[6144, 1024]
blk.49.attn_norm.weight

F32

[6144]
blk.49.attn_output.weight

Q8_0

[6144, 6144]
blk.49.attn_q.weight

Q8_0

[6144, 6144]
blk.49.attn_v.weight

Q8_0

[6144, 1024]
blk.49.ffn_down.weight

Q8_0

[16384, 6144]
blk.49.ffn_gate.weight

Q8_0

[6144, 16384]
blk.49.ffn_norm.weight

F32

[6144]
blk.49.ffn_up.weight

Q8_0

[6144, 16384]

blk.50

blk.50.attn_k.weight

Q8_0

[6144, 1024]
blk.50.attn_norm.weight

F32

[6144]
blk.50.attn_output.weight

Q8_0

[6144, 6144]
blk.50.attn_q.weight

Q8_0

[6144, 6144]
blk.50.attn_v.weight

Q8_0

[6144, 1024]
blk.50.ffn_down.weight

Q8_0

[16384, 6144]
blk.50.ffn_gate.weight

Q8_0

[6144, 16384]
blk.50.ffn_norm.weight

F32

[6144]
blk.50.ffn_up.weight

Q8_0

[6144, 16384]

blk.51

blk.51.attn_k.weight

Q8_0

[6144, 1024]
blk.51.attn_norm.weight

F32

[6144]
blk.51.attn_output.weight

Q8_0

[6144, 6144]
blk.51.attn_q.weight

Q8_0

[6144, 6144]
blk.51.attn_v.weight

Q8_0

[6144, 1024]
blk.51.ffn_down.weight

Q8_0

[16384, 6144]
blk.51.ffn_gate.weight

Q8_0

[6144, 16384]
blk.51.ffn_norm.weight

F32

[6144]
blk.51.ffn_up.weight

Q8_0

[6144, 16384]

blk.52

blk.52.attn_k.weight

Q8_0

[6144, 1024]
blk.52.attn_norm.weight

F32

[6144]
blk.52.attn_output.weight

Q8_0

[6144, 6144]
blk.52.attn_q.weight

Q8_0

[6144, 6144]
blk.52.attn_v.weight

Q8_0

[6144, 1024]
blk.52.ffn_down.weight

Q8_0

[16384, 6144]
blk.52.ffn_gate.weight

Q8_0

[6144, 16384]
blk.52.ffn_norm.weight

F32

[6144]
blk.52.ffn_up.weight

Q8_0

[6144, 16384]

blk.53

blk.53.attn_k.weight

Q8_0

[6144, 1024]
blk.53.attn_norm.weight

F32

[6144]
blk.53.attn_output.weight

Q8_0

[6144, 6144]
blk.53.attn_q.weight

Q8_0

[6144, 6144]
blk.53.attn_v.weight

Q8_0

[6144, 1024]
blk.53.ffn_down.weight

Q8_0

[16384, 6144]
blk.53.ffn_gate.weight

Q8_0

[6144, 16384]
blk.53.ffn_norm.weight

F32

[6144]
blk.53.ffn_up.weight

Q8_0

[6144, 16384]

blk.54

blk.54.attn_k.weight

Q8_0

[6144, 1024]
blk.54.attn_norm.weight

F32

[6144]
blk.54.attn_output.weight

Q8_0

[6144, 6144]
blk.54.attn_q.weight

Q8_0

[6144, 6144]
blk.54.attn_v.weight

Q8_0

[6144, 1024]
blk.54.ffn_down.weight

Q8_0

[16384, 6144]
blk.54.ffn_gate.weight

Q8_0

[6144, 16384]
blk.54.ffn_norm.weight

F32

[6144]
blk.54.ffn_up.weight

Q8_0

[6144, 16384]

blk.55

blk.55.attn_k.weight

Q8_0

[6144, 1024]
blk.55.attn_norm.weight

F32

[6144]
blk.55.attn_output.weight

Q8_0

[6144, 6144]
blk.55.attn_q.weight

Q8_0

[6144, 6144]
blk.55.attn_v.weight

Q8_0

[6144, 1024]
blk.55.ffn_down.weight

Q8_0

[16384, 6144]
blk.55.ffn_gate.weight

Q8_0

[6144, 16384]
blk.55.ffn_norm.weight

F32

[6144]
blk.55.ffn_up.weight

Q8_0

[6144, 16384]

output.weight

Q8_0

[6144, 32768]
output_norm.weight

F32

[6144]